FDI Attack Detection Scheme based on Nonlinear Prediction and Deep Learning

In a smart grid system, an FDI attacker can construct a false data injection attack vector without changing the measurement residual, and traditional bad data detection methods cannot detect such attacks. To overcome such difficult, this paper proposed an effective false data injection attack detection algorithm, in which deep learning algorithms are trained by the measured data and false data, and the appropriate network parameters are obtained to predict the time-varying power data and then he wavelet transform-CNN (Convolutional Neural Networks) is employed for prediction and making detection decision. The experiment under IEEE14 system and IEEE118 system proves the effectiveness of the novel method.